- University of Texas at Arlington, Department of Civil Engineering, United States of America (michelle.hummel@uta.edu)
Exposure to impaired ambient air quality, including fine particulate matter (PM2.5), poses significant risks to human health. Monitoring air pollutants is essential for understanding pollution trends, assessing exposure risks, and informing mitigation strategies. However, traditional regulatory-grade air quality monitoring networks are often sparse and costly to operate, limiting their ability to provide data at high spatial resolution. Low-cost sensors offer an alternative by enabling the deployment of localized monitoring stations with broader coverage, but their accuracy can be compromised under variable environmental conditions. To address this, data fusion techniques can be used to integrate data from multiple sensors and improve air quality predictions. However, integrating multimodal data presents challenges, including incompatible measurement units, spatial and temporal resolutions, and inherent uncertainties.
Here, we propose a probabilistic spatiotemporal model based on the stochastic advection-diffusion (SAD) equation for data fusion in air quality monitoring. The SAD model offers computational efficiency and flexibility, allowing it to handle large datasets while accounting for prediction uncertainties in air quality data. This probabilistic approach is well-suited for air quality managers and policymakers, as it not only predicts air quality with high accuracy but also provides interpretable model parameters that offer insights into the underlying processes driving air pollution. The approach is then evaluated using PM2.5 data from the Coastal Bend Region of Texas, an area facing growing environmental concerns due to expanding industrial development. When the spatiotemporal model is integrated with data from both regulatory-grade stations and low-cost sensors, error is reduced by 40% compared to the nearest regulatory-grade monitor located 20 km away and 11% compared to the nearest low-cost sensor located 1 km away. The model captures 78% of observed data within a 75% confidence interval, demonstrating its ability to accurately represent uncertainty. This method provides a promising framework for integrating diverse air quality data sources, addressing uncertainties, and enhancing community-engaged pollution monitoring efforts.
How to cite: Hummel, M. and Choi, B.: Enhancing Low-Cost Sensor Networks through Multimodal Data Fusion: Application of a Probabilistic Spatiotemporal Air Quality Model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10518, https://doi.org/10.5194/egusphere-egu25-10518, 2025.